Related Experiment Video
Updated: Jan 7, 2026
Using Computer-based Image Analysis to Improve Quantification of Lung Metastasis in the 4T1 Breast Cancer Model
Published on: October 2, 2020
Radiologist- and computer-based chest imaging quantification at 12 months post transplant correlates with baseline
Mikail Malik1, Gauri Rani Karur2, Apoorva Muralidhar2
1Temerty Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada.
Background:
Baseline lung allograft dysfunction (BLAD), defined as failure to achieve ≥ 80% predicted spirometry after lung transplant, is associated with reduced survival. This study aimed to determine the prevalence and character of computed tomography (CT) abnormalities in BLAD.
Methods:
In this retrospective cohort study, we analyzed adult first-time double-lung transplant recipients (12/2017-10/2021) who had 12-month CT chest and concurrent BLAD/non-BLAD status assigned. Three radiologists used a semi-quantitative ordinal score to evaluate ground glass opacities (GGO), reticulation, consolidation, pleural effusion, bronchiectasis, and air-trapping. Machine learning-trained lung texture analysis provided CT radiomic data: CT-measured total lung capacity (CTTLC), pulmonary vessel volume (PVV), GGO, reticulation, and hyperlucency. Parametric response mapping measured functional small airways disease. Receiver operating characteristic analysis and logistic regression identified radiologic features of BLAD.
Results:
BLAD patients (n = 59, 46%) had longer intubation duration, longer index hospitalization post-transplant, and lower donor-to-recipient total lung capacity (TLC) ratio than non-BLAD patients (n = 69, 54%). Radiologist-assessed scoring identified more pleural effusions in BLAD with no significant differences in GGO or air trapping. Computer-aided CT demonstrated more reticulation, GGO, and parenchymal density in BLAD, with no difference in functional small airways disease. CTTLC indexed for height was lower in BLAD, while PVV was higher. PVV was significantly associated with BLAD in univariable analysis (OR 2.30, 95% CI 1.38-3.83, p < 0.001), and remained strong after adjusting for age, sex, and native disease (OR = 2.65,95% CI 1.45-4.84, p = 0.002).
Conclusions:
Computer-aided CT elicited structural changes in BLAD not captured by limited-slice radiologist review. CTTLC and PVV were the strongest radiologic predictors of BLAD, likely reflecting restrictive physiology and vascular remodeling.
